Long‐Term Impact of Minimum Wages on Workers’ Careers: Evidence from Two Decades of Longitudinal Linked Employer–Employee Data

• DOI: http://doi.org/10.1111/sjoe.12327
• Published date: 01 October 2019
• Date: 01 October 2019
• Author: Ana Rute Cardoso

Scand. J. of Economics 121(4), 1337–1380, 2019

DOI: 10.1111/sjoe.12327

Long-Term Impact of Minimum Wages on

Workers’ Careers: Evidence from Two

Decades of Longitudinal Linked

Employer–Employee Data*

Ana Rute Cardoso†

Institute for Economic Analysis (CSIC), 08193 Bellater ra, Barcelona, Spain

anarute.cardoso@iae.csic.es

Abstract

Weanalyze the impact of high youth minimum wages, relyingon two decades of linked employer–

employee data and a major law change. Alternative treatment/control groups follow from two

strands of the literature, one tracking low-skilled workers employed before the law change, who

are eligible for a large wage increase, and one tracking the employmentof full cohor ts, whether

working or in school when the lawchanged. High minimum wages led to a short-term wage gain,

which faded over time. They did not jeopardize employment prospects. Changes in the hours

worked by part-time workers point to increased job attachment.

Keywords: Employment; hours of work; teenage labor markets; wage growth

JEL classification:J08; J31; J24

I. Introduction

There is limited evidence on the impact of exposure to high youth

minimum wages on long-term labor market fortunes. Yet, economic theory

identifies several outcomes that are likely to be influenced by an increase

in the mandatory minimum wage. In particular, employers might reduce

the quantity of labor demanded and undercut the provision of training.

Likewise, individuals might change their labor supply if the minimum

wage influences job attachment or schooling decisions. These changes

would affect the accumulation of human capital and thus labor market

performance over the life cycle. However, depending on the theoretical

setting, we are confronted with contradictory predictions on the operation

*I am grateful to Manuel Arellano, David Card, Jo˜ao Cerejeira, David Dorn, Javier

Fern ´andez Blanco, Dan Hamermesh, David Jaeger, Patrick Kline, Ferran Ma˜n´e, Pedro Portugal,

Jos´e Varej˜ao, and the par ticipants of severalseminars and conferences for helpful comments, as

well as the Statistics Department of the Portuguese Ministry of Employment for data access. I

acknowledge the support of the Spanish Ministry of the Economy (SeveroOchoa Programme for

Centresof Excellence in R&D, SEV20150563), the AEI-Spain, and FEDER-UE (grant ECO2016-

76734-P).

†Also affiliated with Barcelona GSE.

C

The editors of The Scandinavian Journal of Economics 2018.

1338 Long-term impact of minimum wages

of these mechanisms, and the scarcity of empirical evidence limits our

understanding of the process.

This paper analyzes the long-term impact of exposure to high youth

minimum wages on workers’ careers, focusing on the wage profile and

employment – at both intensive and extensive margins – over the life cycle.

Two sets of remarkable conditions are combined for the analysis of

this topic. First, a longitudinal matched employer–employee dataset of

unusual quality is available, which tracks the worker’s employment and

wage history from a young age and over two decades. Specifically, it covers

the population of workers and firms in the private sector in Portugal. The

few existing studies that deal with the long-term impact of the minimum

wage (Neumark and Nizalova, 2007; Dube etal., 2010; Dolton et al., 2012)

could not rely on individual longitudinal data.

Secondly, a sharp change in the legislation took place in the mid-1980s,

leading to a very large wage increase for a very specific group of workers.

On 1 January 1987, the minimum wage for workers aged 17 increased

by 50 percent, and for those aged 18 and 19 it increased by 33 percent.

These legislative changes were exploited by Pereira (2003), who investigated

the impact of minimum wages on firm-level employment, and by Portugal

and Cardoso (2006), who analyzed worker flows and firms’ recruitment

and dismissal policies. Both studies concentrated on firm-level employment

(stocks or flows) and were restricted to the short term, at most two years

after the change in legislation.

We follow up on those studies, extending their analyses into the long

term and for a wider set of outcomes. The core identification strategy

follows from Portugal and Cardoso (2006). Workers aged 16–18 in 1986

were eligible in January 1987 for a major increase in the applicable

minimum wage and were thus defined as the treatment group. Given that

those cohorts were eligible for the high minimum wage at different ages,

for different durations, and to different extents, we will also quantify the

impact of the minimum wage separately on each treated cohort. The control

group consists of workers aged 19 in 1986, who were never exposed to

the high youth minimum wage. In line with Portugal and Cardoso (2006),

we start out by defining the control and treatment groups conditional

on being employed in 1986, a strategy that has both advantages and

limitations. The advantage of this identification strategy is, first of all, that

the members of the treatment group were actually exposed to the high

minimum wage early in their working life. Secondly, it guarantees that

both the treatment and control groups were observed in the labor market,

having already left school, before any change in the law. Hence schooling

is not endogenous. However, this strategy raises concerns about sample

selection bias, as individuals working at young ages (16, 17, or 18) might

differ from those working at the age of 19, potentially leading to different

C

The editors of The Scandinavian Journal of Economics 2018.

A. R. Cardoso 1339

wage and employment patterns over the life cycle. In particular, those who

drop out of school at the age of 16 are most likely to differ from those

who join the labor market later. Therefore, we also perform the analysis by

defining control and treatment groups irrespective of employment status

before the change in legislation. We compare individuals aged 19 and

employed in 1986 (control group) to those aged 19 and employed in 1987,

1988, and 1989 (the treatment groups, potentially exposed to the high

minimum wage for different durations). These two empirical strategies are

in line with different tracks of the previous literature. Whereas the bulk

of the analysis has been aimed at detecting the short-term impact of the

minimum wage on aggregate employment, studies with access to individual

longitudinal data have been aimed at detecting its impact on the probability

of remaining employed for low-skilled individuals already holding a job,

whose wages would have to increase to comply with the new minimum

(Currie and Fallick, 1996; Zavodny, 2000; Stewart, 2004a,b; Stewart

and Swaffield, 2008). Both approaches, whether concentrating on overall

employment for workers potentially affected or on subsequent employment

for low-wage workers who hold a job and are actually affected by the

minimum wage increase, are interesting in themselves. We will explore

both research avenues as we define alternative sets of treatment and control

groups.

We further check whether the schooling achievement of the cohorts

aged 16–19 in 1986 could have been influenced by the change in the law.

This analysis, although constrained by data limitations, suggests that the

sharp rise in youth minimum wages was a contributing factor in the rise

in university completion rates that took place in the country at that time.

This finding reinforces the interest in comparing the treatment and control

groups already observed in the labor market before the change in the law,

whose decisions to leave school had thus not been influenced by the youth

minimum wage level.

Section II describes the mechanisms that can cause youth minimum

wages to have a long-term impact on workers’ careers. Section III explores

the institutional and economic setting in Portugal. In particular, it provides

evidence on changes in the teenage wage distribution after the minimum

wage increase, showing the high degree of compliance with the new

legislation. Section IV describes the data and provides details on the

treatment and control groups. Section V addresses the question of whether

high youth minimum wages jeopardize future wage growth. It devotes

special attention to tackling the problem of separate identification of cohort,

age, and year effects in a wage regression. Section VI addresses the

question of whether high youth minimum wages undermine long-term

employment prospects. It first concentrates on the probability of working

part-time, conditional on working. It then models the number of hours

C

The editors of The Scandinavian Journal of Economics 2018.